AIMS Provenance: ATLAS

Publisher source

The Asteroid Terrestrial-impact Last Alert System (ATLAS) is a global wide-field survey operated by the University of Hawaiʻi to search continuously for hazardous near-Earth objects.

ATLAS products include reduced images, difference images, and detection catalogs. The ATLAS PDS documentation describes .ddc catalogs as difference-image detections produced by the ATLAS pipeline. Their source measurements include position, photometry, PSF shape, duplicate-selection state, and vartest classifier probabilities. The source finder supports fixed, free, trailed, and streaked PSF fits relevant to moving objects.

What AIMS selects

For each ATLAS night, AIMS reads exposure headers and .ddc difference-detection catalogs supplied for the sites with data. It uses FITS headers for exposure timing and filter metadata and .ddc rows for detection measurements.

A row is retained only when all of these are true:

Ptr > 200 or Pkn > 200;
Dup >= 0;
dmag >= 0.

Ptr and Pkn are ATLAS probabilities for an unknown asteroid/transient and a known asteroid, on the publisher’s 0-to-999 scale. Negative Dup values identify measurements rejected in favor of a better nearby detection.

If multiple retained rows produce the same AIMS observation id, AIMS deterministically prefers the stronger moving-source fit using detection class, elongation, magnitude uncertainty, classifier confidence, and stable source order.

Coverage and observatories

AIMS coverage starts 2015-06-18 and is continuing.

ATLAS prefix	Site	AIMS/MPC observatory code
`01a`	Mauna Loa	`T08`
`02a`	Haleakalā	`T05`
`03a`	South Africa	`M22`
`04a`	Chile	`W68`

Normalization into AIMS

AIMS field	ATLAS source or derivation
`observation_id`	Deterministic string from exposure id, RA, Dec, x, and y.
`exposure_id`	FITS `OBSID`.
`dataset_id`	Constant `atlas`.
`observatory_code`	Derived from the ATLAS site prefix.
Observation time	FITS `MJD-OBS + EXPTIME / 2 / 86400`.
`filter`	FITS `FILTER`.
`ra_deg`, `dec_deg`	`.ddc` `RA`, `Dec`.
`mag`, `mag_sigma`	`.ddc` `mag`, `dmag`.
`ra_sigma_deg`, `dec_sigma_deg`	AIMS approximation from PSF size and photometric SNR.
`healpixel`	Derived from RA/Dec at nside=128 NESTED.

Astrometric uncertainty approximation

The selected .ddc path does not supply explicit sky-coordinate sigmas, so AIMS currently computes:

fwhm_proxy = sqrt(major^2 + minor^2)
snr = 1.0857 / dmag
sigma = (fwhm_proxy / snr) / 3600 degrees

Rows with dmag == 0 are assigned dmag = 0.001 only to keep this calculation finite. That floor implies very high signal-to-noise and a very small sigma; it is not conservative uncertainty inflation.

The calculation uses source major and minor values without a camera-specific pixel-scale conversion or local WCS Jacobian, copies one scalar into both axes, and retains no cross-axis correlation. It is a legacy matching stand-in, not a calibrated sky-plane covariance.

Published AIMS products

In the AIMS BigQuery dataset, normalized ATLAS rows are written to aims_atlas, copied into aims_detections_hot, and exposed through the aims_detections_unified view with dataset_id = atlas.

The same normalized rows are exported to Google Cloud Storage through the daily aims_detections_parquet product. AIMS derives aims_frames_unified BigQuery rows and aims_frames_parquet search indexes from ATLAS exposure ids and spatial cells. Complete detection/frame pairs are listed in aims_unified_parquet_catalog/catalog.json for consumers such as Precovery. The exported rows remain filtered .ddc measurements; the Parquet step adds no recalibration.

Scientific validity and limitations

Publisher positions, photometry, classifier values, and duplicate states make the selected rows useful for moving-object search, but AIMS includes only a filtered subset of ATLAS difference detections.
The Ptr/Pkn threshold is an AIMS selection threshold, not a final object classification or a measured completeness boundary.
AIMS follows the publisher duplicate-selection signal and adds a deterministic same-id tie-break. That resolves representational duplicates; it does not establish that nearby distinct rows are the same physical observation.
The astrometric uncertainty approximation is not scientifically calibrated for camera geometry, distinct east/north axes, correlation, or cross-survey comparison. Consumers needing calibrated covariance should not use it as such.
Public PDS documentation describes the source products, while operational AIMS delivery may use ATLAS-provided paths rather than the public archive. The scientific row semantics remain those of the supplied headers and .ddc catalogs.